Rheumatism rehabilitation intelligent evaluation system based on ICF idea and classification system

Abstract

The invention relates to the technical field of medical rehabilitation information, in particular to a rheumatism rehabilitation intelligent evaluation system based on an ICF concept and a classification system, and the system comprises a multi-source data module which is used for collecting and integrating multi-dimensional data; the dynamic evaluation module is used for constructing a dynamic weight distribution model based on an ICF theory and updating an evaluation result and a radar map according to the multi-dimensional data; the data feature module is used for performing high-dimensional feature dimension reduction processing according to the evaluation result and the radar map on the basis of a potential profile analysis algorithm, and identifying a hidden subtype; and the causal intervention module is used for nesting multi-dimensional dynamic coupling evaluation of an ICF theory, subtype identification of subtype analysis and causal modeling of a structural equation model to be applied to rheumatism rehabilitation decision making based on the hidden subtype when the rheumatism rehabilitation decision making system is used. The problems that traditional evaluation dimensions are single, data are statically isolated and are difficult to adapt to individual differences are solved, and the accuracy of an intervention strategy and the rehabilitation effect of a patient are improved.

Description

Technical Field

[0001] The present invention relates to the field of medical rehabilitation information technology, and in particular to an intelligent assessment system for rheumatism rehabilitation based on the ICF concept and classification system. Background Art

[0002] The International Classification of Functioning, Disability and Health (ICF) is a classification system published by the World Health Organization. It provides a standardized language and framework for describing health status outcomes. It comprehensively assesses an individual's health status from three levels: body function and structure, activity, and participation, rather than being limited to the disease itself. The application of the ICF framework in rheumatic rehabilitation assessment can more comprehensively and accurately reflect the patient's degree of functional impairment, the impact on daily living activities, and the problems faced in social participation.

[0003] The patent publication number is CN119274800A, which states in its specification that "the present invention relates to the field of medical rehabilitation technology, and in particular to a remote rehabilitation intelligent evaluation system. The system includes a rehabilitation monitoring module, a remote service module, an interactive feedback module and an intelligent evaluation module; the rehabilitation monitoring module includes an emotional signal monitoring module, a physiological signal monitoring module and a gamification performance monitoring module for monitoring the patient's emotional condition, physiological condition and reaction condition; the remote service module connects the rehabilitation monitoring module, the interactive feedback module and the intelligent evaluation module; the interactive feedback module is used for information interaction and feedback between the patient, the rehabilitation therapist and the remote service module; the intelligent evaluation module is used for the remote service module to monitor the patient's emotional condition, physiological condition and reaction condition. The rehabilitation index data and optimized rehabilitation information obtained through module processing are analyzed and evaluated. The present invention realizes remote, comprehensive and refined intelligent evaluation of the patient's rehabilitation status by comprehensively monitoring the patient's emotional, physiological and reaction status. Although the above technology is centered on multi-source data fusion + dynamic weight allocation, and achieves the purpose of precise and personalized intelligent rehabilitation management through collaborative evaluation of the three dimensions of emotion / physiology / reaction, combined with gamification testing and adaptive algorithms, the traditional evaluation system is difficult to take into account the dynamic coupling of the patient's multidimensional health status and group heterogeneity. In the evaluation and decision-making process, the evaluation dimensions are single, the data are static and isolated, and it is difficult to adapt to individual differences, resulting in rough intervention strategies and low effects.

[0004] In summary, the development of an intelligent assessment system for rheumatic disease rehabilitation based on the ICF concept and classification system remains a key issue that needs to be urgently addressed in the field of medical rehabilitation information technology. Summary of the Invention

[0005] The purpose of this invention is to solve the problems in the existing technology that the traditional assessment system is difficult to take into account the dynamic coupling and group heterogeneity of the patient's multidimensional health status, and the evaluation dimensions are single, the data are static and isolated, and it is difficult to adapt to individual differences in the assessment and decision-making process, resulting in rough intervention strategies and low effects.

[0006] To achieve the above objectives, the present invention provides an intelligent assessment system for rheumatic disease rehabilitation based on the ICF concept and classification system, comprising: Multi-source data module, used to collect and integrate multi-dimensional data; A dynamic evaluation module, based on ICF theory, constructs a dynamic weight allocation model and updates the evaluation results and radar chart according to the multi-dimensional data; A data feature module, based on a latent profile analysis algorithm, performs high-dimensional feature dimensionality reduction processing according to the evaluation results and radar chart to identify latent subtypes; The causal intervention module uses the structural equation model to construct a causal path network based on the latent subtype and combines it with knowledge graph technology to generate a targeted intervention strategy library; A closed-loop feedback module is used to establish an evaluation-decision-making bidirectional feedback mechanism, reversely input the dynamic evaluation module to optimize weight distribution, and iteratively update the causal path network.

[0007] Beneficial effects Compared with the known public technology, the technical solution provided by the present invention has the following beneficial effects: When used, the present invention nests the multi-dimensional dynamic coupling evaluation of ICF theory, the subtype identification of latent profile analysis, and the causal modeling of the structural equation model and applies them to rheumatic disease rehabilitation decision-making, which is convenient for solving the problems of traditional evaluation with a single dimension, static and isolated data, and difficulty in adapting to individual differences, and is conducive to improving the accuracy of intervention strategies and patient rehabilitation effects.

[0008] When used, the present invention learns and optimizes causal intervention pathways for different patient subtypes, making the pushed rehabilitation plans more personalized and effective. Patients upload their daily status at home through smart terminals, and the system updates the weights and causal models in real time. Doctors adjust treatment strategies accordingly, forming an adaptive and intelligent rehabilitation closed loop, which is conducive to improving rehabilitation effects and patient satisfaction and reducing waste of medical resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 This is a system diagram of the intelligent assessment system for rheumatic disease rehabilitation based on the ICF concept and classification system of the present invention.

[0010] Figure 2 This is the software interface diagram of the rheumatism rehabilitation intelligent assessment system based on the ICF concept and classification system of the present invention.

[0011] Figure 3 This is a diagram of the completion interface based on ICF patient precision assessment.

[0012] Figure 4 This is a status interface diagram based on ICF patient accurate assessment. DETAILED DESCRIPTION

[0013] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0014] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units that are not explicitly listed or are inherent to these processes, methods, products or devices.

[0015] See Figure 1-4 , this application further describes the present invention in detail with reference to the accompanying drawings: Example: like Figure 1 As shown, the present invention provides an intelligent assessment system for rheumatic disease rehabilitation based on the ICF concept and classification system, including: Multi-source data module, used to collect and integrate multi-dimensional data; Furthermore, the operation process of the multi-source data module includes: The system is used to collect and integrate multi-dimensional data. Through the API interface protocol and the hospital information system, image archiving and communication system, the patient's biochemical examination data, imaging data and medication records are collected in real time. At the same time, a patient-side interactive interface is developed to support the entry of self-assessment data based on the five dimensions of the ICF framework, including disease activity, physical function, quality of life, social support, and work efficiency. The patient-side App interface supports the completion of the ICF five-dimensional questionnaire and generates the following self-assessment data vector:

[0016] Where, Represents a five-dimensional evaluation vector under the ICF framework, represents a set with five elements, Indicates any The evaluation scores of the dimensions, The score range for each dimension is from 0 to 1, where 0 represents extremely poor and 1 represents excellent. All scales are normalized and mapped to the interval [0, 1]. DS evidence theory is introduced to construct a data fusion algorithm. For data conflict scenarios, a manual review process is initiated through outlier cross-validation rules, providing respective trust allocations for the two data sources. The synthesized evidence is:

[0017] Where, Indicates that for the proposition The combined trust distribution function value is, Represents the first subset of evidence sources The trust distribution function value of Represents the second subset of evidence sources The trust distribution function value of Indicates that in all In combination, the intersection is equal to All pairs Find the sum on top, represents the normalization factor for all non-conflicting parts, Represents all subset combinations when all intersections are empty sets Sum, Representation subset and The intersection of is empty, extending to When there are multiple data sources, build a fusion process:

[0018] Where, represents the fused trust distribution function, It represents a method of fusing multiple pieces of evidence using the combination rule of Dempster-Shafer evidence theory. Indicates the 1st to The original evidence trust distribution function.

[0019] Specifically, the multi-source data module of this system collects patients' biochemical examination data, medical imaging data and medication records in real time through API interfaces with medical information platforms such as hospital information systems, image archiving and communication systems, ensuring the comprehensiveness and timeliness of clinical data. At the same time, a patient-side App is developed, and a self-assessment questionnaire with five dimensions (disease activity, physical function, quality of life, social support, and work efficiency) is designed based on the ICF theoretical framework. Patients can directly enter their subjective feelings to generate a standardized five-dimensional evaluation vector. The scale score range is uniformly mapped to the interval of 0 to 1, which facilitates unified analysis, enhances the integrity and multi-dimensional understanding of the information, and realizes the effective combination of subjective and objective data based on the standardized patient self-assessment scale. The introduction of evidence theory improves the reliability and robustness of data fusion. Combined with the manual review mechanism, it facilitates the timely identification and processing of abnormal data.

[0020] A dynamic evaluation module, based on ICF theory, constructs a dynamic weight allocation model and updates the evaluation results and radar chart according to the multi-dimensional data; Furthermore, the operation process of the dynamic assessment module includes: Based on the ICF theory, the five dimensions are decomposed into a set of quantifiable sub-indicators, and a dynamic weight allocation model is constructed. The initial weight of one of the current dimensions is set to be , and when fluctuations in the relevant sub-indicators are detected, a nonlinear update is triggered:

[0021] Where, Indicates at a point in time When, Dynamic weight values ​​of dimensions, represents a natural constant, represents the adjustable sensitivity coefficient, Indicates time When, The change in dimension, Five ICF core dimensions Do the sum, Represents the exponential weighted value of each dimension. Based on the multi-dimensional data, the Bayesian network algorithm is used to establish the causal relationship between dimensions and construct a directed acyclic graph , where It is a directed acyclic graph, with nodes is the ICF dimension, edge express yes For any node , define the conditional probability distribution, expression:

[0022] Where, represents the probability distribution, Indicates the evaluation dimensions, express The parent node set of Indicates that when the parent node is known The conditional probability distribution of Represents a function parameter used to model the conditional probability The input variables are , represents a probability model, Indicates the The parameters of a function.

[0023] Furthermore, the operation process of the dynamic assessment module includes: Calculate the final weighted assessment score , each dimension The value of can be obtained by summarizing the sub-indicators and taking the expected value, the expression is:

[0024] Where, Indicates the The overall evaluation score of the moment, Indicates that from Dimensions are accumulated to Dimension, Indicates the Dimensions at the moment The weight of Indicates the Dimensions at the moment The specific score of Indicates that from Sub-indicators are added to There are a total of Sub-indicators, Indicates the Dimension The weight of each sub-indicator satisfies , Indicates the In the dimension, Sub-indicators at time The rating is based on five dimensions. , automatically draw a five-dimensional radar chart, and store the historical evaluation sequence after each round of evaluation , where Indicates time A collection of historical evaluation data, Indicates time The historical assessment data collection is used to visualize the assessment trends and update the assessment results and radar charts.

[0025] Specifically, this system continuously monitors five dimensions of physical function, mobility, participation, environmental factors, and psychological state for neurorehabilitation patients, dynamically adjusts the weights of various indicators, and accurately reflects subtle changes in the patient's rehabilitation process. The Bayesian network reveals the causal paths between different dimensions, helps identify key influencing factors, and implements targeted rehabilitation interventions. The dynamic weight allocation mechanism improves the sensitivity and adaptability of the assessment, and the Bayesian network causal modeling enhances the scientific nature and reliability of the assessment.

[0026] A data feature module, based on a latent profile analysis algorithm, performs high-dimensional feature dimensionality reduction processing according to the evaluation results and radar chart to identify latent subtypes; Furthermore, the operation process of the data feature module includes: The latent profile analysis algorithm is based on the evaluation results and radar chart to perform high-dimensional feature dimensionality reduction processing, the expression is:

[0027] Where, Indicates the The feature vector of the sample The probability density value of Indicates the The feature vector of the sample, Indicates that all The sum of the number of hidden subtypes, represents the number of latent subtypes in the Gaussian mixture model, Indicates the The weights of the Gaussian components, Indicates the Gaussian distribution for samples The probability density function value of Indicates the The mean vector of a Gaussian distribution, Indicates the The covariance matrix of a Gaussian distribution is used to determine the optimal number of patient subtype classifications using the minimum information entropy criterion. The expression is:

[0028] Where, is the entropy regularized objective function to measure the number of hidden subtypes given The comprehensive index of model division clarity and complexity under For all samples arrive The sum of Indicates the numbering of all hidden subtypes arrive The sum of Indicates the The samples belong to The membership of the hidden subtype, Indicates the The samples belong to The logarithmic function of the membership of hidden subtypes, represents the hyperparameter, Indicates the number of feature dimensions of each sample, represents the number of parameters of the covariance matrix of each Gaussian component, represents the number of parameters for a single hidden subtype, Indicates all The total number of parameters required for hidden subtypes, represents the logarithm of the sample size, represents the optimal number of hidden subtypes, Indicates taking the minimum objective function value, Indicates the number of cryptic subtypes in arrive Perform traversal search within the range.

[0029] Furthermore, the operation process of the data feature module includes: Through iterative clustering, a multi-round iterative strategy based on potential response weight adjustment is adopted, the expression is:

[0030] Where, Indicates the The sample in The feature representation vector at the round iteration, Indicates the The sample in The new feature representation after rounds of iteration, represents the optimal number of hidden subtypes, Indicates the The sample in In the round of iteration, The membership of the hidden subtype, Indicates the The occult subtype The center representation vector during round iteration, Indicates the Samples to Update direction of hidden subtypes, is the step length, identifying the cryptic subtype.

[0031] Specifically, based on the latent profile analysis algorithm, this system performs dimensionality reduction processing and latent subtype identification on the high-dimensional features of the patient's multidimensional evaluation results and radar chart data. For rheumatic disease rehabilitation management, it processes multidimensional data including disease indicators, quality of life scores and functional assessments, and performs fine grouping of patients to identify different pathological subtypes, such as inflammation-dominant type and functional disorder type, thereby providing a scientific basis for subsequent personalized treatment and intervention, achieving effective dimensionality reduction and reasonable grouping of high-dimensional data. Multi-round iterative clustering strategies enhance the stability and recognition accuracy of this system. Identifying latent subtypes helps to reveal the heterogeneity of patient groups, support the formulation of personalized medical plans, and improve clinical diagnosis and treatment effects and patient management levels.

[0032] The causal intervention module uses the structural equation model to construct a causal path network based on the latent subtype and combines it with knowledge graph technology to generate a targeted intervention strategy library; Furthermore, the operational process of the causal intervention module includes: Based on the latent subtype, a causal path network was constructed using the structural equation model, and the latent variable vector was set as , the observed variable vector is , the causal path network, expression:

[0033] Where, represents the observed variable vector, represents the latent variable vector, Indicates the influence strength of each latent variable on the observed variable, represents the observation error, represents the influence of one latent variable on another latent variable. represents the influence matrix of exogenous variables on latent variables, Indicates unexplained random fluctuations.

[0034] Furthermore, the operational process of the causal intervention module includes: The conduction paths between dimensions are verified by the maximum likelihood estimation method, the covariance structure and log-likelihood function are constructed, and then the negative log-likelihood is minimized by gradient descent to obtain the optimal parameters. The expression is:

[0035] Where, Represents the parameter vector function, Indicates the influence strength of each latent variable on the observed variable, represents the identity matrix, represents the influence of one latent variable on another latent variable. represents the propagation matrix of causal effects in the system, represents the covariance matrix between latent variables, The inverse transposed matrix of the propagation matrix representing the causal effects in the system, yes The transpose of represents the measurement error covariance matrix, Indicates that the parameter The goodness of fit of the model under Represents the parameter vector The logarithmic function of the absolute value of Indicates the The number of samples with hidden subtypes, represents the scaling factor, Indicates the The sample covariance matrix of the latent subtypes, represents the inverse matrix of the covariance matrix, represents the sum of all elements on the main diagonal, represents the optimal parameter vector, Represents the parameters that minimize the objective function At the same time, combined with knowledge graph technology, a targeted intervention strategy library is generated and a triple knowledge graph is constructed. The expression is:

[0036] Where, represents a collection of knowledge graphs, Represents a knowledge edge in the graph, automatically matching personalized decision-making plans for patients with different latent subtypes.

[0037] Specifically, this system constructs a causal path network based on the structural equation model of latent subtypes, accurately depicts the relationship between latent variables and observed variables and the causal transmission mechanism between latent variables, and identifies different latent subtypes in the patient population for rheumatic disease rehabilitation patients, such as immune response type, inflammation-dominant type, etc., uses the structural equation model to reveal the causal relationship between various pathological indicators, and combines clinical knowledge graphs to generate personalized drug and rehabilitation intervention plans, realizing the construction of a knowledge-driven targeted intervention strategy library, enhancing the scientificity and practicality of decision-making, automatically matching personalized plans, improving treatment effects and patient satisfaction, and promoting the development of precision medicine.

[0038] A closed-loop feedback module is used to establish an evaluation-decision-making bidirectional feedback mechanism, reversely input the dynamic evaluation module to optimize weight distribution, and iteratively update the causal path network; Furthermore, the operation process of the closed-loop feedback module includes: The method for establishing an evaluation-decision-making bidirectional feedback mechanism is reversely input into the dynamic evaluation module. After the decision plan is implemented, the intervention effect data fed back by the patient is introduced into the dynamic loss function, which is reversely input into the dynamic evaluation module. The dynamic weight is iteratively optimized using the mirror descent method with entropy constraints. The expression is:

[0039] Where, Indicates the The feedback loss function at each time point is: The square of the Euclidean distance is used to measure the gap between the predicted value and the true feedback value. Indicates the The patient's real feedback vector at each time point, Indicates the The system predicts the output vector at each time point, It indicates that the hyperparameter is used to adjust the weight of the subsequent entropy regularization term in the total loss. Indicates the current weight distribution The information entropy of Indicates at a point in time Time Dynamic weight values ​​of dimensions, represents the entropy function, Indicates in After the update The weight value of each dimension, Indicates that the right side is the value before normalization. represents the natural exponential function, represents the learning rate, Represents the loss function for the current The partial derivatives of the dimension weights.

[0040] Furthermore, the operation process of the closed-loop feedback module includes: Iteratively update the causal pathway network by using a policy gradient-based reinforcement learning algorithm to obtain a pathway selection strategy based on the patient's long-term evaluation results and radar chart ,expression:

[0041] Where, Indicates the The patient's real feedback vector at each time point, Indicates the The system predicts the output vector at each time point, Representation Strategy The total expected return, Indicates that in the strategy The expected value under Indicates from arrive The summation symbol, Represents the discount factor Power, Indicates the discount factor value , The smaller the difference between the evaluation value and the actual feedback, the higher the reward. represents the penalty term coefficient, The Kullback-Leibler divergence measures the difference between two probability distributions. Indicates the current time The causal path network constructed, Indicates the last moment The causal path network of The greater the deviation between the current path network and the previous path structure, the higher the penalty.

[0042] Specifically, this system realizes the process of dynamically optimizing patients' personalized rehabilitation plans by building a two-way feedback mechanism of evaluation and decision-making. In the rehabilitation management of patients with rheumatic diseases, through the implementation of multiple intervention plans and the dynamic input of patient feedback data, the weights of evaluation dimensions, such as social support and physical function, are continuously adjusted to reflect the patient's latest health status. At the same time, this system learns and optimizes causal intervention paths for different patient subtypes, making the pushed rehabilitation plans more personalized and effective. Patients upload their daily status through smart terminals at home, and this system updates the weights and causal models in real time. Doctors adjust treatment strategies accordingly to form an adaptive and intelligent rehabilitation closed loop, which is conducive to improving rehabilitation effects and patient satisfaction and reducing waste of medical resources.

[0043] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. The intelligent assessment system for rheumatic disease rehabilitation based on the ICF concept and classification system is characterized by: include: Multi-source data module, used to collect and integrate multi-dimensional data; A dynamic evaluation module, based on ICF theory, constructs a dynamic weight allocation model and updates the evaluation results and radar chart according to the multi-dimensional data; A data feature module, based on a latent profile analysis algorithm, performs high-dimensional feature dimensionality reduction processing according to the evaluation results and radar chart to identify latent subtypes; The causal intervention module uses the structural equation model to construct a causal path network based on the latent subtype and combines it with knowledge graph technology to generate a targeted intervention strategy library; A closed-loop feedback module is used to establish an evaluation-decision-making bidirectional feedback mechanism, reversely input the dynamic evaluation module to optimize weight distribution, and iteratively update the causal path network.

2. The intelligent assessment system for rheumatic disease rehabilitation based on the ICF concept and classification system according to claim 1 is characterized in that: The operation process of the multi-source data module includes: The system is used to collect and integrate multi-dimensional data. Through the API interface protocol and the hospital information system, image archiving and communication system, the patient's biochemical examination data, imaging data and medication records are collected in real time. At the same time, a patient-side interactive interface is developed to support the entry of self-assessment data based on the five dimensions of the ICF framework, including disease activity, physical function, quality of life, social support, and work efficiency. The patient-side App interface supports the completion of the ICF five-dimensional questionnaire and generates the following self-assessment data vector: Where, Represents a five-dimensional evaluation vector under the ICF framework, represents a set with five elements, Indicates any The evaluation scores of the dimensions, The score range for each dimension is from 0 to 1, where 0 represents extremely poor and 1 represents excellent. All scales are normalized and mapped to the interval [0, 1]. DS evidence theory is introduced to construct a data fusion algorithm. For data conflict scenarios, a manual review process is initiated through outlier cross-validation rules, providing respective trust allocations for the two data sources. The synthesized evidence is: Where, Indicates that for the proposition The combined trust distribution function value is, Represents the first subset of evidence sources The trust distribution function value of Represents the second subset of evidence sources The trust distribution function value of Indicates that in all In combination, the intersection is equal to All pairs Find the sum on top, represents the normalization factor for all non-conflicting parts, Represents all subset combinations when all intersections are empty sets Sum, Representation subset and The intersection of is empty, extending to When there are multiple data sources, build a fusion process: Where, represents the fused trust distribution function, It represents a method of fusing multiple pieces of evidence using the combination rule of Dempster-Shafer evidence theory. Indicates the 1st to The original evidence trust distribution function.

3. The intelligent assessment system for rheumatic disease rehabilitation based on the ICF concept and classification system according to claim 2 is characterized in that: The operation process of the dynamic assessment module includes: Based on the ICF theory, the five dimensions are decomposed into a set of quantifiable sub-indicators, and a dynamic weight allocation model is constructed. The initial weight of one of the current dimensions is set to be , and when fluctuations in the relevant sub-indicators are detected, a nonlinear update is triggered: Where, Indicates at a point in time When, Dynamic weight values ​​of dimensions, represents a natural constant, represents the adjustable sensitivity coefficient, Indicates time When, The change in dimension, Five ICF core dimensions Do the sum, Represents the exponential weighted value of each dimension. Based on the multi-dimensional data, the Bayesian network algorithm is used to establish the causal relationship between dimensions and construct a directed acyclic graph , where It is a directed acyclic graph, with nodes is the ICF dimension, edge express yes For any node , define the conditional probability distribution, expression: Where, represents the probability distribution, Indicates the evaluation dimensions, express The parent node set of Indicates that when the parent node is known The conditional probability distribution of Represents a function parameter used to model the conditional probability The input variables are , represents a probability model, Indicates the The parameters of a function.

4. The intelligent assessment system for rheumatic disease rehabilitation based on the ICF concept and classification system according to claim 3 is characterized in that: The operation process of the dynamic assessment module includes: Calculate the final weighted assessment score , each dimension The value of can be obtained by summarizing the sub-indicators and taking the expected value, the expression is: Where, Indicates the The overall evaluation score of the moment, Indicates that from Dimensions are accumulated to Dimension, Indicates the Dimensions at the moment The weight of Indicates the Dimensions at the moment The specific score of Indicates that from Sub-indicators are added to There are a total of Sub-indicators, Indicates the Dimension The weight of each sub-indicator satisfies , Indicates the In the dimension, Sub-indicators at time The rating is based on five dimensions. , automatically draw a five-dimensional radar chart, and store the historical evaluation sequence after each round of evaluation , where Indicates time A collection of historical evaluation data, Indicates time The historical assessment data collection is used to visualize the assessment trends and update the assessment results and radar charts.

5. The intelligent assessment system for rheumatic disease rehabilitation based on the ICF concept and classification system according to claim 4 is characterized in that: The operation process of the data feature module includes: The latent profile analysis algorithm is based on the evaluation results and radar chart to perform high-dimensional feature dimensionality reduction processing, the expression is: Where, Indicates the The feature vector of the sample The probability density value of Indicates the The feature vector of the sample, Indicates that all The sum of the number of hidden subtypes, represents the number of latent subtypes in the Gaussian mixture model, Indicates the The weights of the Gaussian components, Indicates the Gaussian distribution for samples The probability density function value of Indicates the The mean vector of a Gaussian distribution, Indicates the The covariance matrix of a Gaussian distribution is used to determine the optimal number of patient subtype classifications using the minimum information entropy criterion. The expression is: Where, is the entropy regularized objective function to measure the number of hidden subtypes given The comprehensive index of model division clarity and complexity under For all samples arrive The sum of Indicates the numbering of all hidden subtypes arrive The sum of Indicates the The samples belong to The membership of the hidden subtype, Indicates the The samples belong to The logarithmic function of the membership of hidden subtypes, represents the hyperparameter, Indicates the number of feature dimensions of each sample, represents the number of parameters of the covariance matrix of each Gaussian component, represents the number of parameters for a single hidden subtype, Indicates all The total number of parameters required for hidden subtypes, represents the logarithm of the sample size, represents the optimal number of hidden subtypes, Indicates taking the minimum objective function value, Indicates the number of cryptic subtypes in arrive Perform traversal search within the range.

6. The intelligent assessment system for rheumatic disease rehabilitation based on the ICF concept and classification system according to claim 5 is characterized in that: The operation process of the data feature module includes: Through iterative clustering, a multi-round iterative strategy based on potential response weight adjustment is adopted, the expression is: Where, Indicates the The sample in The feature representation vector at the round iteration, Indicates the The sample in The new feature representation after rounds of iteration, represents the optimal number of hidden subtypes, Indicates the The sample in In the round of iteration, The membership of the hidden subtype, Indicates the The occult subtype The center representation vector during round iteration, Indicates the Samples to Update direction of hidden subtypes, is the step length, identifying the cryptic subtype.

7. The intelligent assessment system for rheumatic disease rehabilitation based on the ICF concept and classification system according to claim 6 is characterized in that: The operational process of the causal intervention module includes: Based on the latent subtype, a causal path network was constructed using the structural equation model, and the latent variable vector was set as , the observed variable vector is , the causal path network, expression: Where, represents the observed variable vector, represents the latent variable vector, Indicates the influence strength of each latent variable on the observed variable, represents the observation error, represents the influence of one latent variable on another latent variable. represents the influence matrix of exogenous variables on latent variables, Indicates unexplained random fluctuations.

8. The intelligent assessment system for rheumatic disease rehabilitation based on the ICF concept and classification system according to claim 7 is characterized in that: The operational process of the causal intervention module includes: The conduction paths between dimensions are verified by the maximum likelihood estimation method, the covariance structure and log-likelihood function are constructed, and then the negative log-likelihood is minimized by gradient descent to obtain the optimal parameters. The expression is: Where, Represents the parameter vector function, Indicates the influence strength of each latent variable on the observed variable, represents the identity matrix, represents the influence of one latent variable on another latent variable. represents the propagation matrix of causal effects in the system, represents the covariance matrix between latent variables, The inverse transposed matrix of the propagation matrix representing the causal effects in the system, yes The transpose of represents the measurement error covariance matrix, Indicates that the parameter The goodness of fit of the model under Represents the parameter vector The logarithmic function of the absolute value of Indicates the The number of samples with hidden subtypes, represents the scaling factor, Indicates the The sample covariance matrix of the latent subtypes, represents the inverse matrix of the covariance matrix, represents the sum of all elements on the main diagonal, represents the optimal parameter vector, Represents the parameters that minimize the objective function At the same time, combined with knowledge graph technology, a targeted intervention strategy library is generated and a triple knowledge graph is constructed. The expression is: Where, represents a collection of knowledge graphs, Represents a knowledge edge in the graph, automatically matching personalized decision-making plans for patients with different latent subtypes.

9. The intelligent assessment system for rheumatic disease rehabilitation based on the ICF concept and classification system according to claim 8 is characterized in that: The operation process of the closed-loop feedback module includes: The method for establishing an evaluation-decision-making bidirectional feedback mechanism is reversely input into the dynamic evaluation module. After the decision plan is implemented, the intervention effect data fed back by the patient is introduced into the dynamic loss function, which is reversely input into the dynamic evaluation module. The dynamic weight is iteratively optimized using the mirror descent method with entropy constraints. The expression is: Where, Indicates the The feedback loss function at each time point is: The square of the Euclidean distance is used to measure the gap between the predicted value and the true feedback value. Indicates the The patient's real feedback vector at each time point, Indicates the The system predicts the output vector at each time point, It indicates that the hyperparameter is used to adjust the weight of the subsequent entropy regularization term in the total loss. Indicates the current weight distribution The information entropy of Indicates at a point in time Time Dynamic weight values ​​of dimensions, represents the entropy function, Indicates in After the update The weight value of each dimension, Indicates that the right side is the value before normalization. represents the natural exponential function, represents the learning rate, Represents the loss function for the current The partial derivatives of the dimension weights.

10. The intelligent assessment system for rheumatic disease rehabilitation based on the ICF concept and classification system according to claim 9 is characterized in that: The operation process of the closed-loop feedback module includes: Iteratively update the causal pathway network by using a policy gradient-based reinforcement learning algorithm to obtain a pathway selection strategy based on the patient's long-term evaluation results and radar chart ,expression: Where, Indicates the The patient's real feedback vector at each time point, Indicates the The system predicts the output vector at each time point, Representation Strategy The total expected return, Indicates that in the strategy The expected value under Indicates from arrive The summation symbol, Represents the discount factor Power, Indicates the discount factor value , The smaller the difference between the evaluation value and the actual feedback, the higher the reward. represents the penalty term coefficient, The Kullback-Leibler divergence measures the difference between two probability distributions. Indicates the current time The causal path network constructed, Indicates the last moment The causal path network of The greater the deviation between the current path network and the previous path structure, the higher the penalty.

Images